pith. sign in

arxiv: 2601.10397 · v3 · pith:LQ7WDS7Onew · submitted 2026-01-15 · 🧬 q-bio.NC

Reshaping Neural Representation via Associative, Presynaptic Short-Term Plasticity

Genki Shimizu , Taro Toyoizumi This is my paper

Pith reviewed 2026-05-21 16:40 UTC · model grok-4.3

classification 🧬 q-bio.NC
keywords short-term plasticityassociative STPTsodyks-MarkramFisher informationrelease probabilitytemporal codingneural representationinformation theory
0
0 comments X

The pith

Associative short-term plasticity derives from information maximization and splits into postsynaptic firing tracking and presynaptic onset detection.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a normative theory for associative short-term synaptic plasticity by extending Fisher-information learning to Tsodyks-Markram synapses. It derives learning rules for baseline synaptic weight and release probability that maximize stimulus information while respecting resource constraints. These rules naturally separate into a postsynaptic term that follows local firing rates and a presynaptic term that is phase-advanced to detect the beginning of stimuli. A sympathetic reader would care because this provides a functional explanation for how synapses can rapidly reshape neural representations to support temporal aspects of coding, such as onset sensitivity and replay dynamics, linking theory to experimental observations of associative STP.

Core claim

By extending Fisher-information-based learning to Tsodyks-Markram synapses, learning rules are derived for baseline weight and release probability that maximize stimulus information under resource constraints. The rules consist of a postsynaptic term tracking local firing and a presynaptic, phase-advanced term that selectively detects stimulus onset. This onset sensitivity favors anti-causal connectivity for slowly varying inputs and produces frequency-dependent phase selectivity in linear response analysis.

What carries the argument

The derived learning rules for release probability in the Tsodyks-Markram synapse model, which separate into a postsynaptic local-activity term and a presynaptic phase-advanced onset-detection term.

If this is right

  • Onset sensitivity in the presynaptic term favors anti-causal connectivity for slowly varying inputs.
  • Response offset is enhanced during drive and reverse replay occurs after drive removal in recurrent circuits.
  • STP produces frequency-dependent phase selectivity in neural responses.
  • Constraints on release probability tune the temporal asymmetry of the coding.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This could enable rapid reconfiguration of temporal codes in response to changing stimulus statistics without long-term weight changes.
  • Experiments could test if presynaptic plasticity rules match the predicted phase advance for onset detection in sensory pathways.
  • The framework connects information optimization to observed associative effects in short-term plasticity across brain regions.

Load-bearing premise

Short-term plasticity serves to maximize information about stimuli subject to constraints on synaptic resources.

What would settle it

An experiment showing that changes in presynaptic release probability do not preferentially detect stimulus onsets in a phase-advanced manner when pre- and postsynaptic activity are paired would falsify the derived rules.

Figures

Figures reproduced from arXiv: 2601.10397 by Genki Shimizu, Taro Toyoizumi.

Figure 3
Figure 3. Figure 3: 25 [PITH_FULL_IMAGE:figures/full_fig_p025_3.png] view at source ↗
read the original abstract

Short-term synaptic plasticity (STP) is often regarded as a presynaptic filter of spikes, independent of postsynaptic activity. Recent experiments, however, indicate an associative STP that depends on pre- and postsynaptic coactivation. We develop a normative, information-theoretic theory of associative STP. Extending Fisher-information-based learning to Tsodyks-Markram synapses, we derive learning rules for baseline weight and release probability that maximize stimulus information under resource constraints. The rules split into a postsynaptic term tracking local firing and a presynaptic, phase-advanced term that selectively detects stimulus onset. For slowly varying inputs, this onset sensitivity favors anti-causal connectivity and enhances response offset during drive and reverse replay after drive removal in recurrent circuits. Linear-response analysis shows that STP yields frequency-dependent phase selectivity and that release-probability constraints tune temporal asymmetry. These results identify release-probability plasticity as a principled substrate for rapidly reconfigurable temporal coding.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript develops a normative, information-theoretic theory of associative short-term synaptic plasticity (STP) in Tsodyks-Markram synapses. Extending Fisher-information maximization under resource constraints, it derives separate learning rules for baseline synaptic weight (a postsynaptic term tracking local firing) and release probability (a presynaptic, phase-advanced term selective for stimulus onset). For slowly varying inputs the theory predicts anti-causal connectivity, offset enhancement during drive, and reverse replay after drive removal in recurrent circuits. Linear-response analysis is used to show frequency-dependent phase selectivity whose temporal asymmetry is tuned by release-probability constraints.

Significance. If the central derivations are valid, the work supplies a principled, parameter-light account of why presynaptic STP can be associative and how it supports rapid temporal reconfigurability in neural circuits. The explicit link between information maximization, the split into pre- and postsynaptic terms, and concrete circuit-level predictions (anti-causal wiring, reverse replay) is a clear strength. The absence of direct empirical validation and the reliance on linearization for the key onset-detection claim limit immediate impact, but the framework is falsifiable and could guide future experiments on STP.

major comments (2)
  1. [§3] §3 (derivation of presynaptic rule): the phase-advanced presynaptic term for release probability is obtained only after linearizing the Tsodyks-Markram mean and variance of postsynaptic current around a steady state. The manuscript does not report a direct comparison between this linearized Fisher gradient and the full nonlinear stochastic response for the slowly varying inputs that are claimed to produce anti-causal connectivity and offset enhancement; any mismatch would undermine the selective onset detection and the reported frequency-dependent phase selectivity.
  2. [Linear-response analysis] Linear-response analysis (likely §4): the frequency-dependent phase selectivity and the tuning of temporal asymmetry by release-probability constraints rest on the same linear approximation. Without numerical checks showing that higher-order state fluctuations (u, x) do not erase the phase advance in the relevant regime, the central claim that release-probability plasticity is a substrate for rapidly reconfigurable temporal coding remains provisional.
minor comments (2)
  1. [Abstract / Introduction] The resource constraints that bound the optimization are stated only qualitatively in the abstract and introduction; an explicit statement of the constraint functional (e.g., Eq. (X)) would clarify the derivation for readers.
  2. [Methods] Notation for the Tsodyks-Markram variables (u, x, etc.) and the precise definition of the postsynaptic current should be introduced with a single reference equation before the linear-response calculations begin.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. The concerns focus on the reliance on linearization in the derivations and analyses. We address each point below and have performed additional numerical validations to confirm the robustness of the approximations in the relevant regimes. These will be incorporated into the revised manuscript.

read point-by-point responses

  1. Referee: [§3] §3 (derivation of presynaptic rule): the phase-advanced presynaptic term for release probability is obtained only after linearizing the Tsodyks-Markram mean and variance of postsynaptic current around a steady state. The manuscript does not report a direct comparison between this linearized Fisher gradient and the full nonlinear stochastic response for the slowly varying inputs that are claimed to produce anti-causal connectivity and offset enhancement; any mismatch would undermine the selective onset detection and the reported frequency-dependent phase selectivity.

    Authors: We acknowledge that the presynaptic rule is derived via linearization of the Tsodyks-Markram mean and variance around steady state, which is appropriate for the slowly varying inputs considered. To directly address potential discrepancies, we have now run numerical simulations of the full nonlinear stochastic Tsodyks-Markram model driven by the same slowly varying inputs and compared the resulting Fisher gradients to the linearized predictions. The comparisons show that the phase advance, onset selectivity, and anti-causal connectivity predictions are preserved with only small quantitative deviations. A new supplementary figure and accompanying text will be added to document these checks. revision: yes

  2. Referee: [Linear-response analysis] Linear-response analysis (likely §4): the frequency-dependent phase selectivity and the tuning of temporal asymmetry by release-probability constraints rest on the same linear approximation. Without numerical checks showing that higher-order state fluctuations (u, x) do not erase the phase advance in the relevant regime, the central claim that release-probability plasticity is a substrate for rapidly reconfigurable temporal coding remains provisional.

    Authors: The linear-response analysis yields closed-form expressions for frequency-dependent phase selectivity and its tuning by release-probability constraints. We agree that explicit checks against higher-order fluctuations are valuable. Accordingly, we have simulated the full nonlinear recurrent network dynamics (including stochastic evolution of u and x) under the derived plasticity rules across the relevant frequency range. These simulations confirm that the phase advance and temporal asymmetry persist and are not erased by state fluctuations within the operating regime of the model. We will add these numerical results and a brief discussion of the approximation limits to the revised manuscript. revision: yes

Circularity Check

0 steps flagged

Derivation from external Fisher-information principle is self-contained with no reduction to inputs

full rationale

The paper extends an established information-maximization objective to the Tsodyks-Markram synapse model and derives learning rules for baseline weight and release probability. The split into a postsynaptic term and a presynaptic phase-advanced term follows directly from applying the Fisher information gradient to the model's mean and variance under resource constraints, without any quoted step that defines a quantity in terms of itself or renames a fitted parameter as a prediction. No self-citation is invoked as the load-bearing justification for the central premise, and the approach remains normative rather than tautological. The provided abstract and context give no equations showing that any reported outcome (onset detection, anti-causal connectivity) reduces by construction to the resource constraints or to a prior self-citation chain.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review performed on abstract only; the central modeling choice is the information-maximization objective applied to an existing STP model, with no explicit free parameters or new entities named in the abstract.

axioms (1)
  • domain assumption Maximizing stimulus information under resource constraints is the appropriate objective for deriving short-term plasticity rules
    This objective is invoked to obtain the learning rules for baseline weight and release probability.

pith-pipeline@v0.9.0 · 5682 in / 1279 out tokens · 64596 ms · 2026-05-21T16:40:48.865484+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages

  1. [1]

    T. V. Bliss, T. Lomo, Long-Lasting Potentiation of Synaptic Transmission in the Dentate Area of the Anaesthetized Rabbit Following Stimulation of the Perforant Path.The Journal of Physiology232(2), 331–356 (1973), doi:10.1113/jphysiol.1973.sp010273

    work page doi:10.1113/jphysiol.1973.sp010273 1973

  2. [2]

    Bi, M.-m

    G.-q. Bi, M.-m. Poo, Synaptic Modifications in Cultured Hippocampal Neurons: Dependence on Spike Timing, Synaptic Strength, and Postsynaptic Cell Type.The Journal of Neuroscience 18(24), 10464 (1998), doi:10.1523/JNEUROSCI.18-24-10464.1998

    work page doi:10.1523/jneurosci.18-24-10464.1998 1998

  3. [3]

    R. S. Zucker, W. G. Regehr, Short-Term Synaptic Plasticity.Annual Review of Physiology64, 355–405 (2002), doi:10.1146/annurev.physiol.64.092501.114547

    work page doi:10.1146/annurev.physiol.64.092501.114547 2002

  4. [4]

    Ucar,et al., Mechanical Actions of Dendritic-Spine Enlargement on Presynaptic Exocytosis

    H. Ucar,et al., Mechanical Actions of Dendritic-Spine Enlargement on Presynaptic Exocytosis. Nature600(7890), 686–689 (2021), doi:10.1038/s41586-021-04125-7

    work page doi:10.1038/s41586-021-04125-7 2021

  5. [5]

    Kasai, H

    H. Kasai, H. Ucar, Y. Morimoto, F. Eto, H. Okazaki, Mechanical Transmission at Spine Synapses: Short-term Potentiation and Working Memory.Current Opinion in Neurobiology 80, 102706 (2023), doi:10.1016/j.conb.2023.102706

    work page doi:10.1016/j.conb.2023.102706 2023

  6. [6]

    Linsker, Self-Organization in a Perceptual Network.Computer21(3), 105–117 (1988), doi:10.1109/2.36

    R. Linsker, Self-Organization in a Perceptual Network.Computer21(3), 105–117 (1988), doi:10.1109/2.36

    work page doi:10.1109/2.36 1988

  7. [7]

    Toyoizumi, J.-P

    T. Toyoizumi, J.-P. Pfister, K. Aihara, W. Gerstner, Generalized Bienenstock-Cooper-Munro Rule for Spiking Neurons That Maximizes Information Transmission.Proceedings of the National Academy of Sciences of the United States of America102(14), 5239–44 (2005), doi:10.1073/pnas.0500495102

    work page doi:10.1073/pnas.0500495102 2005

  8. [8]

    Toyoizumi, J.-p

    T. Toyoizumi, J.-p. Pfister, K. Aihara, W. Gerstner, Spike-Timing Dependent Plasticity and Mu- tual Information Maximization for a Spiking Neuron Model, inAdvances in Neural Information Processing Systems(MIT Press), vol. 17 (2005), pp. 1409–1416

    work page 2005

  9. [9]

    Toyoizumi, K

    T. Toyoizumi, K. Aihara, S. I. Amari, Fisher Information for Spike-Based Population Decoding. Physical Review Letters97(9) (2006), doi:10.1103/PhysRevLett.97.098102. 30

    work page doi:10.1103/physrevlett.97.098102 2006

  10. [10]

    Rotman, P.-Y

    Z. Rotman, P.-Y. Deng, V. A. Klyachko, Short-Term Plasticity Optimizes Synaptic Information Transmission.The Journal of Neuroscience: The Official Journal of the Society for Neuro- science31(41), 14800–14809 (2011), doi:10.1523/JNEUROSCI.3231-11.2011

    work page doi:10.1523/jneurosci.3231-11.2011 2011

  11. [11]

    W. G. Regehr, Short-Term Presynaptic Plasticity.Cold Spring Harbor Perspectives in Biology 4(7), a005702 (2012), doi:10.1101/cshperspect.a005702

    work page doi:10.1101/cshperspect.a005702 2012

  12. [12]

    Dayan, L

    P. Dayan, L. F. Abbott,Theoretical Neuroscience: Computational and Mathematical Modeling of Neural Systems, Computational Neuroscience (Massachusetts Institute of Technology Press, Cambridge, Mass) (2001)

    work page 2001

  13. [13]

    Brunel, J

    N. Brunel, J. P. Nadal, Mutual Information, Fisher Information, and Population Coding.Neural Computation10(7), 1731–1757 (1998), doi:10.1162/089976698300017115

    work page doi:10.1162/089976698300017115 1998

  14. [14]

    H. S. Seung, H. Sompolinsky, Simple Models for Reading Neuronal Population Codes.Pro- ceedings of the National Academy of Sciences of the United States of America90(22), 10749– 10753 (1993), doi:10.1073/pnas.90.22.10749

    work page doi:10.1073/pnas.90.22.10749 1993

  15. [15]

    M. V. Tsodyks, H. Markram, The Neural Code between Neocortical Pyramidal Neurons De- pends on Neurotransmitter Release Probability.Proceedings of the National Academy of Sci- ences of the United States of America94(2), 719–723 (1997), doi:10.1073/pnas.94.2.719

    work page doi:10.1073/pnas.94.2.719 1997

  16. [16]

    Tsodyks, K

    M. Tsodyks, K. Pawelzik, H. Markram, Neural Networks with Dynamic Synapses.Neural Computation10(4), 821–835 (1998), doi:10.1162/089976698300017502

    work page doi:10.1162/089976698300017502 1998

  17. [17]

    T. Haga, T. Fukai, Recurrent Network Model for Learning Goal-Directed Sequences through Reverse Replay.eLife7, e34171 (2018), doi:10.7554/eLife.34171

    work page doi:10.7554/elife.34171 2018

  18. [18]

    R. C. Malenka, M. F. Bear, LTP and LTD: An Embarrassment of Riches.Neuron44(1), 5–21 (2004), doi:10.1016/j.neuron.2004.09.012

    work page doi:10.1016/j.neuron.2004.09.012 2004

  19. [19]

    S. Fusi, P. J. Drew, L. F. Abbott, Cascade Models of Synaptically Stored Memories.Neuron 45(4), 599–611 (2005), doi:10.1016/j.neuron.2005.02.001. 31

    work page doi:10.1016/j.neuron.2005.02.001 2005

  20. [20]

    Zenke, E

    F. Zenke, E. J. Agnes, W. Gerstner, Diverse Synaptic Plasticity Mechanisms Orchestrated to Form and Retrieve Memories in Spiking Neural Networks.Nature Communications6(1), 6922 (2015), doi:10.1038/ncomms7922

    work page doi:10.1038/ncomms7922 2015

  21. [21]

    L. F. Abbott, W. G. Regehr, Synaptic Computation.Nature431(7010), 796–803 (2004), doi:10.1038/nature03010

    work page doi:10.1038/nature03010 2004

  22. [22]

    E. S. Fortune, G. J. Rose, Short-Term Synaptic Plasticity as a Temporal Filter.Trends in Neurosciences24(7), 381–385 (2001), doi:10.1016/s0166-2236(00)01835-x

    work page doi:10.1016/s0166-2236(00)01835-x 2001

  23. [23]

    M. E. Hasselmo, The Role of Acetylcholine in Learning and Memory.Current Opinion in Neurobiology16(6), 710–715 (2006), doi:10.1016/j.conb.2006.09.002

    work page doi:10.1016/j.conb.2006.09.002 2006

  24. [24]

    S. J. Sara, The Locus Coeruleus and Noradrenergic Modulation of Cognition.Nature Reviews Neuroscience10(3), 211–223 (2009), doi:10.1038/nrn2573

    work page doi:10.1038/nrn2573 2009

  25. [25]

    Wu,et al., Presynaptic Release Probability Determines the Need for Sleep (2025), doi: 10.1101/2025.10.16.682770

    Y. Wu,et al., Presynaptic Release Probability Determines the Need for Sleep (2025), doi: 10.1101/2025.10.16.682770

    work page doi:10.1101/2025.10.16.682770 2025

  26. [26]

    Huang, C

    S. Huang, C. Piao, C. B. Beuschel, T. G ¨otz, S. J. Sigrist, Presynaptic Active Zone Plasticity Encodes Sleep Need in Drosophila.Current Biology30(6), 1077–1091.e5 (2020), doi:10. 1016/j.cub.2020.01.019

    work page 2020

  27. [27]

    D. J. Foster, M. A. Wilson, Reverse Replay of Behavioural Sequences in Hippocampal Place Cells during the Awake State.Nature440(7084), 680–683 (2006), doi:10.1038/nature04587

    work page doi:10.1038/nature04587 2006

  28. [28]

    K. Diba, G. Buzs ´aki, Forward and Reverse Hippocampal Place-Cell Sequences during Ripples. Nature Neuroscience10(10), 1241–1242 (2007), doi:10.1038/nn1961

    work page doi:10.1038/nn1961 2007

  29. [29]

    R. E. Ambrose, B. E. Pfeiffer, D. J. Foster, Reverse Replay of Hippocampal Place Cells Is Uniquely Modulated by Changing Reward.Neuron91(5), 1124–1136 (2016), doi:10.1016/j. neuron.2016.07.047

    work page doi:10.1016/j 2016

  30. [30]

    Buzs ´aki, Hippocampal Sharp Wave-ripple: A Cognitive Biomarker for Episodic Memory and Planning.Hippocampus25(10), 1073–1188 (2015), doi:10.1002/hipo.22488

    G. Buzs ´aki, Hippocampal Sharp Wave-ripple: A Cognitive Biomarker for Episodic Memory and Planning.Hippocampus25(10), 1073–1188 (2015), doi:10.1002/hipo.22488. 32

    work page doi:10.1002/hipo.22488 2015

  31. [31]

    H. F. ´Olafsd´ottir, D. Bush, C. Barry, The Role of Hippocampal Replay in Memory and Planning. Current Biology28(1), R37–R50 (2018), doi:10.1016/j.cub.2017.10.073

    work page doi:10.1016/j.cub.2017.10.073 2018

  32. [32]

    S. Gais, J. Born, Declarative Memory Consolidation: Mechanisms Acting during Human Sleep.Learning & Memory (Cold Spring Harbor, N.Y.)11(6), 679–685 (2004), doi:10.1101/ lm.80504

    work page 2004

  33. [33]

    Rasch, J

    B. Rasch, J. Born, About Sleep’s Role in Memory.Physiological Reviews93(2), 681–766 (2013), doi:10.1152/physrev.00032.2012

    work page doi:10.1152/physrev.00032.2012 2013

  34. [34]

    M. R. Mehta, M. C. Quirk, M. A. Wilson, Experience-Dependent Asymmetric Shape of Hippocampal Receptive Fields.Neuron25(3), 707–715 (2000), doi:10.1016/s0896-6273(00) 81072-7

    work page doi:10.1016/s0896-6273(00 2000

  35. [35]

    Huxter, N

    J. Huxter, N. Burgess, J. O’Keefe, Independent Rate and Temporal Coding in Hippocampal Pyramidal Cells.Nature425(6960), 828–832 (2003), doi:10.1038/nature02058

    work page doi:10.1038/nature02058 2003

  36. [36]

    Schwalger, M

    T. Schwalger, M. Deger, W. Gerstner, Towards a Theory of Cortical Columns: From Spiking Neurons to Interacting Neural Populations of Finite Size.PLoS Computational Biology13(4), e1005507 (2017), doi:10.1371/journal.pcbi.1005507

    work page doi:10.1371/journal.pcbi.1005507 2017

  37. [37]

    G. K. Ocker, A. Litwin-Kumar, B. Doiron, Self-Organization of Microcircuits in Networks of Spiking Neurons with Plastic Synapses.PLOS Computational Biology11(8), e1004458 (2015), doi:10.1371/journal.pcbi.1004458

    work page doi:10.1371/journal.pcbi.1004458 2015

  38. [38]

    G. K. Ocker, B. Doiron, Training and Spontaneous Reinforcement of Neuronal Assemblies by Spike Timing Plasticity.Cerebral Cortex29(3), 937–951 (2019), doi:10.1093/cercor/bhy001

    work page doi:10.1093/cercor/bhy001 2019

  39. [39]

    Humble, K

    J. Humble, K. Hiratsuka, H. Kasai, T. Toyoizumi, Intrinsic Spine Dynamics Are Critical for Recurrent Network Learning in Models With and Without Autism Spectrum Disorder.Frontiers in Computational Neuroscience13(2019), doi:10.3389/fncom.2019.00038

    work page doi:10.3389/fncom.2019.00038 2019

  40. [40]

    Reyes,et al., Target-Cell-Specific Facilitation and Depression in Neocortical Circuits.Nature Neuroscience1(4), 279–285 (1998), doi:10.1038/1092

    A. Reyes,et al., Target-Cell-Specific Facilitation and Depression in Neocortical Circuits.Nature Neuroscience1(4), 279–285 (1998), doi:10.1038/1092. 33

    work page doi:10.1038/1092 1998

  41. [41]

    Y. Ma, H. Hu, A. Agmon, Short-Term Plasticity of Unitary Inhibitory-to-Inhibitory Synapses Depends on the Presynaptic Interneuron Subtype.The Journal of Neuroscience: The Official Journal of the Society for Neuroscience32(3), 983–988 (2012), doi:10.1523/JNEUROSCI. 5007-11.2012

    work page doi:10.1523/jneurosci 2012

  42. [42]

    Pfister, T

    J.-P. Pfister, T. Toyoizumi, D. Barber, W. Gerstner, Optimal Spike-Timing-Dependent Plasticity for Precise Action Potential Firing in Supervised Learning.Neural Computation18(6), 1318– 1348 (2006), doi:10.1162/neco.2006.18.6.1318

    work page doi:10.1162/neco.2006.18.6.1318 2006

  43. [43]

    B. A. Olshausen, D. J. Field, Emergence of Simple-Cell Receptive Field Properties by Learning a Sparse Code for Natural Images.Nature381(6583), 607–609 (1996), doi:10.1038/381607a0

    work page doi:10.1038/381607a0 1996

  44. [44]

    R. P. N. Rao, D. H. Ballard, Predictive Coding in the Visual Cortex: A Functional Interpretation of Some Extra-Classical Receptive-Field Effects.Nature Neuroscience2(1), 79–87 (1999), doi:10.1038/4580. Acknowledgments The author gratefully acknowledges Kensuke Yoshida and all members of the Toyoizumi Lab for insightful advice and stimulating discussions. ...

    work page doi:10.1038/4580 1999